Qwen3.6 27B
on NVIDIA RTX 3090 · llama.cpp · 100,000 ctx
Use cases
coding
Summary
User reports MTP (Multi-Token Prediction) GGUF with the am17an commit for faster inference.
Setup uses a Q4_0 KV cache and a speculative draft of 2.
At 100k context, VRAM usage is about 19 GB, and performance degrades above 90k context.